JetBrains Air Is Now in Your IDE: What Developers Must Know

JetBrains released Air in IDEs in Early Access on October 1, 2026, shipping its agentic orchestration layer as a free plugin inside IntelliJ, Rider, PhpStorm, WebStorm, and PyCharm rather than as a separate product. Air ships with no agents installed and connects agents that support the Agent Client Protocol (ACP), the Apache-licensed standard JetBrains co-built with Zed and which JetBrains, Zed, Google, GitHub, and 25+ agents have adopted, letting developers run multiple agent sessions in parallel in isolated Git worktrees or Docker containers and review each output as a diff before it lands. The plugin is free, with JetBrains AI credits starting at $10/month for AI Pro if developers use JetBrains' hosted model access instead of their own Anthropic, OpenAI, or Google API keys.

JetBrains just shipped its agentic orchestration layer into every IDE it builds — not as a new product you have to migrate to, but as a plugin inside the IDE you already use. Air in IDEs landed in Early Access on October 1 https://blog.jetbrains.com/ai/2026/10/air-in-ides-eap/ . If you run IntelliJ, Rider, PhpStorm, WebStorm, or PyCharm, you can install it today, connect Claude Agent or Codex or Junie, and run multiple agent sessions in parallel on separate tasks — all without leaving your editor. This is JetBrains’ answer to Cursor. And it’s a more serious answer than most people expected. What Air in IDEs Actually Does Air in IDEs is not an AI model. It ships with no agents installed. Think of it as the orchestration layer: you connect the agents you already use, and Air manages them inside your IDE. The core workflow looks like this: you assign tasks to multiple agents — Codex on the API refactor, Claude Agent on test coverage, Junie debugging the flaky CI test — and they run concurrently in isolated Git worktrees or Docker containers. Each session is tracked from a single view: activity, changed files, outgoing commits, and cost. When an agent finishes, its output shows up as a reviewable diff inside the IDE — same tooling you’d use to review a pull request. You can leave line-level comments for the agent to act on before any changes land in your codebase. That last part matters. Agent output is not auto-applied. You review it. The IDE doesn’t become an autonomous pipeline you watch from a distance — you stay in control of what lands. The ACP Angle: JetBrains Is Playing Protocol Politics Air works with agents that support the Agent Client Protocol ACP — an open standard JetBrains co-built with Zed, released under the Apache license. The easiest way to understand ACP: it’s the LSP moment for coding agents. The Language Server Protocol let any editor support any language through a single shared standard. ACP does the same for agents — any editor that implements ACP can connect to any agent that implements ACP, without bespoke integrations for every combination. It uses JSON-RPC 2.0 over stdin/stdout, and it’s already adopted by JetBrains, Zed, Google, GitHub, and 25+ agents https://www.jetbrains.com/acp/ . JetBrains and Zed also launched an ACP Registry — a discoverable directory of compatible agents built directly into the IDE. This is not a coincidence. JetBrains co-authoring the protocol that defines how agents talk to editors is a strategic move. They’re not just shipping an AI feature — they’re positioning themselves as infrastructure. The Part That Cursor Can’t Do Here’s what’s genuinely different about Air versus terminal-first agent tools like Claude Code or standalone Codex: Air-connected agents can invoke IDE tooling as skills. An agent can trigger a debug run to investigate a failing test. It can run the profiler. It has access to the refactoring engine with full, multi-file codebase context. Agents that can debug are not the same as agents that can only read and write files. When an agent can profile a slow function and see the call stack, it’s working with the same context a senior developer would have. That depth has always been JetBrains’ edge — 26 years of IDE tooling. Air is the first time that tooling is available to agents. How to Get It Two paths depending on your IDE version: - IDE 2026.2 and earlier: Settings → Plugins → Marketplace → search “Air” → Install - IDE 2026.3 EAP: Already bundled. Settings → Plugins → Bundled → enable “Air” The Air plugin on JetBrains Marketplace https://plugins.jetbrains.com/plugin/33314-air is free. You connect your own agent — Claude, Codex, Gemini CLI, GitHub Copilot, Junie, or OpenCode. If you already have API keys for Anthropic, OpenAI, or Google, your cost from JetBrains is zero. JetBrains AI credits are available if you want to use their hosted model access, starting at $10/month for AI Pro. Windows support launched alongside the EAP with 30 free AI credits for new users. Why JetBrains Needed This Cursor hit $2B ARR with over a million paying users by turning VS Code into an AI-first editor. Claude Code and Codex run agent pipelines from terminals, outside any IDE entirely. JetBrains abandoned Fleet — its attempt at a modern editor — in late 2024. The trajectory was clear: if JetBrains didn’t bring agents into IntelliJ and Rider, those IDEs were going to get left behind as developers shifted their workflows to agentic tools. Air in IDEs is the counter-move. Don’t compete on editor UX against VS Code. Don’t compete on models against Anthropic or OpenAI. Compete on the one thing JetBrains has that no one else does https://thenewstack.io/jetbrains-air-agents-ide/ : 26 years of deep language intelligence, debugger integration, and static analysis that agents can now actually use. It’s in Early Access. It will have rough edges. But if you’re already running a JetBrains IDE and curious about multi-agent development, there’s no reason not to install the plugin today.